positive feedback trading and stock return autocorrelation: the
Post on 12-Feb-2022
2 Views
Preview:
TRANSCRIPT
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
1
Positive Feedback Trading and Stock Return
Autocorrelation: The Case of Morocco
A. Charafi and M. Achkir
School of Business Administration, Al Akhawayn University, PO Box 104, Hassan II Avenue, Ifrane 53000 Morocco
Corresponding Author E-mail: a.charafi@aui.ma
Abstract. This paper investigates the presence of positive feedback trading in the
Casablanca stock exchange and measures the profitability and the effectiveness of
selected herding strategies. The MADEX returns from 2004 to 2010 are analyzed,
modeled, and forecasted for that purpose using linear autoregressive models, GARCH
processes, and E-GARCH processes. Relying on the Sentana and Wadhwani’s positive
feedback model, this paper explores the link between feedback trading, serial
autocorrelations, and volatility. It presents supporting evidence on the persistence of serial
autocorrelations in the index returns suggesting the prevailing influence of feedback
trading activity on return dynamics. The signaling-based simulation results reveal that herd
trading dominates the simple buy and hold strategy and the smart money investors’
strategy both on daily and weekly bases. The results also unveil the impact of the day of
trade on weekly trading outcomes, volatilities, and Sharpe ratios.
Key words: positive feedback trading, autocorrelation, GARCH.
1. Introduction
Throughout the last decade, there has been growing interest in the emerging and
developing stock exchanges and the investment opportunities present in the countries that
host them. These markets are referred to as “frontier markets”; and are characterized by
small market capitalizations, low liquidity levels, and imminent privatization trends. In
addition, these economies are distinguished by the constrained impact of international
events on their prosperity and progress. The reason for this is that the stock exchanges in
such countries list local companies that have limited ties at the international level. They are
also hosted in countries where the restrictive regulation confines the impact of external
shocks and crises on the overall economy. Thus, the stock returns in these markets often
exhibit negative correlations with the more developed ones. These various factors highlight
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
2
risk reduction prospects that attract long term investors that are interested in the
diversification opportunities offered by such economic environments.
The attractiveness of the frontier markets sheds light on their well-functioning and raises
concerns about their efficiency. The question whether the prices incorporate information
instantly and correctly or whether the returns exhibit a random walk or not is the prime
concern of agents involved in these markets. Answers to these questions would give an
indication about the predictability of the markets and about the manner news releases
disseminate into stock prices. Unveiling the nature of the trading conducts in those
markets would also allow a better grasp of how the prices are influenced and how they are
impacted by the various trading strategies.
In a broad sense, one can consider the trading conduct as being governed by two schools;
the fundamental analysis and the technical analysis. The fundamental analysis’ adepts
engage in trading with a solid knowledge of the companies’ standing and in-depth studies
of financial information and industry settings. They are referred to as “rational investors” or
“smart money investors” (SMI) since they rely on forecasting techniques and valuation
models drawn from historical data incorporating various economic factors. The second
method of trading; the technical analysis, is based on examining prices and volumes of
strictly historical data. The practitioners of this type of trading are called trend chasers, or
rational speculators. They trade with the ultimate conviction that the trends’ history repeats
itself and are believed to influence the market movements.
One of the possible ways in which the technical analysts’ behavior could affect share
prices is through feedback trading. Feedback traders are trend followers who base their
strategy on price movements. Positive (negative) feedback traders buy (sell) when prices
rise and sell (buy) when prices fall (rise). Hence, if their presence is significant in a market
they can induce serial autocorrelations of the returns. Such behavior was first documented
by Cutler, Poterba, & Summers (1990) as they revealed the significant presence of serial
autocorrelation in US stock returns. The empirical conclusions that were reached by the
aforementioned authors were a breakthrough in behavioral finance and triggered the
curiosity of other scholars that were more interested in developing a theoretical framework
for such market interactions.
The theoretical aspect of the herding behavior was first dealt with in Shiller’s (1984) work
when he documented evidence of overreaction following dividend announcements in the
US and attributed this phenomenon to social trends. Shiller (1984) builds a model to
capture such behavior which was improved upon by Sentana and Wadhwani (1992) to
become the positive feedback trading model. Empirical studies based on that model
succeed in finding evidence of positive feedback trading in developed markets as well as
in emerging markets. Koutmos,1997; and Koutmos and Saidi, 2001 among others show
how positive feedback traders induce negative autocorrelations in returns in the US and
emerging Asian markets respectively.
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
3
This paper complements the positive feedback trading studies by investigating the
presence and effectiveness of positive feedback trading in the Casablanca stock exchange
(CSE). It also compares trading gains of feedback trading to other chosen strategies.
Daily data of the Casablanca stock exchange over the 2001-2010 period is used for that
purpose. The theoretical facet of this paper is based on the model developed by Shiller
(1984) and Sentana and Wadhwani (1992) and the adopted methodology is grounded on a
GARCH mean model.
2. Motivation and Purpose
There has been no formal empirical study of the positive feedback traders’ activity in the
Moroccan stock market although several analysts report reversion to the mean
phenomena and describe herding effects. Squalli (2006) explains how Colorado’s IPO was
subject to a trend effect in the first weeks of trading. The stock price was expected to
increase, so investors rushed to buy the IPO in masses. After weeks of trading activity and
unfounded price increases, the stocks suffered a series of declines dragging it back to its
fundamental value. Other analysts studying the general market trends relate the CSE
boom to rational speculation. Drissi El Bouzidi (2006) declares that the overall market is
overvalued and that behavioral aspects keep the prices artificially higher than they should
be. After periods where the market could gain 30% in a matter of weeks, the CSE indices
entered a period a repetitive declines in 2009. The market suffered a “psychological crisis”
driven by the small investors’ panic as pointed out by Nhaili (2009). The aforementioned
market dynamics are significant indicators of the prevailing presence of positive feedback
trading. Lack of literature documenting market occurrences leaves analysts, investors, and
scholars with a poor understanding of the effects taking place during different trading
phases.
Our main objective is to investigate evidence of positive feedback trading activity in the
CSE and measure its intensity during market ups and downs. We establish a relationship
between positive feedback trading and the presence of serial autocorrelation in the returns
of the CSE drawing a link between the level of volatility and the nature of trading. We also
examine the effectiveness of various trading strategies in the CSE for several scenarios,
comparing the annualized returns over a seven years period (2004-2010) for daily trading
versus weekly trading and for short selling possibility and no short selling conduct. We
study the profitability prospects for selected trading strategies and techniques; passive
strategy, smart money investor’s strategy, feedback trading, simple hybrid strategy, and
complex hybrid strategy and how they are affected by the day of execution.
3. The Positive Feedback Trading Model
3.1. Positive Feedback Traders
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
4
The main assumption of the model developed by Shiller (1984) and Sentana and Wadhwani (1992) is that the market is mainly composed of two types of agents. The first group is named “smart money investors” or “expected utility maximizers (S). This type of investors relies mostly on the fundamentals corresponding to shares such as profitability, leverage, or cash flows, and its behavior is mainly characterized by risk aversion. The second group of investors, the positive feedback traders also referred to as trend chasers (F) primary trade on price movement and evolution. The demand for the first group is established by the Dynamic Capital Asset Pricing Model developed by Merton (1973) and is given by the following Equation:
�� = ������� �� (1)
Where �� is the fraction of shares demanded by the smart money investors at period t, E��� is the expectation operator calculated from all available information at period t-1 and calculated as the average yearly return assuming the investor buys at the beginning of the year and sells at the end of the same year; R� the rate of return at period t using closing prices; α is the rate at which the demand for shares by the smart money investors is null. Setting α equal to the risk free rate, Equation (1) becomes equivalent to the Dynamic Capital Asset Pricing Model developed by Merton (1973). μ� is the volatility measure as a
function of the conditional variance, μ� = μσ�²) is the conditional variance measuring risk at time t. To account for the risk aversion of rational investors μ′σ��) ≥ 0 so the higher the volatility, the lower the proportion of shares demanded by the smart money investors. If all shares are held by Smart Money investors so that S� = 1 Equation (1) simply becomes the standard Capital Asset Pricing Model (CAPM): E���R�� − α = μσ��� (2)
The Demand for the second group; the feedback traders, is set by the following equation:
F� = �ρ�R���ifR��� ≥ 0ρ�R���ifR��� < 0 (3)
Where ρ�and ρ� indicate the nature of the feedback trading withρ�, ρ� > 0 to capture
positive feedback trading. In the opposite case whereρ�, ρ� < 0, there is negative
feedback trading; i.e. selling (buying) when prices increase (decrease). This demand
equation is more general than the one suggested by Sentana and Wadhwani (1992) where
F� = ρR���.
The equilibrium F� + S� = 1 where all shares are held by both types of investors results in
the equation:
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
5
E���R�� = �α + μσ��� − ρ�μσ���R���ifR��� ≥ 0α + μσ��� − ρ�μσ���R���ifR��� < 0 (4)
The difference between equations (2) and (4) is the additional terms that introduce the
positive feedback traders into the CAPM equation and allows for negative serial
correlation. This equation shows the relationship between the positive feedback trading
and the returns. The terms ρ�μσ���and ρ�μσ��� induce negative autocorrelation between
the index returns at period t-1 and the returns at period t. These terms also show that the
higher the volatilityμσ���, the more negative the autocorrelation.
The rational expectations assumption states that the expectations of crowds would affect
market movements and eventually concretize. This would allow for R� =E���R�� +ε�
which results in the following equation:
R� = �α + μσ��� − ρ�μσ���R��� + εifR��� ≥ 0α + μσ��� − ρ�μσ���R��� + εifR��� < 0 (5)
For testing purposes, equation (5) needs to be transformed into a linear form. The linear
form is more suitable for a regression equation. This conversion is achieved by setting
ρ�μσ��� = ρ'� + ρ��σ�� andρ�μσ��� = ρ'� + ρ��σ��. As a result, equation (5) becomes
R� = (α + λσ�� − ρ'� + ρ��σ���R��� + εifR��� ≥ 0α + λσ�� − ρ'� + ρ��σ���R��� + εifR��� < 0 (6)
In equation (6), the direct impact of feedback traders is given byρ'�) andρ'�) for the case
of low volatility levels. As risk increases, the terms ρ��σ�� and ρ��σ��will have more influence
on the return and the impact of feedback trading will be determined byρ��andρ��. If
ρ'�, ρ'�, ρ��, ρ��< 0, it is an indication that negative feedback traders are more active in the
market. This outcome is more likely to occur during periods of low volatility knowing that
negative feedback trading is only one of the hypotheses.
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
6
3.2. Non-Synchronous Trading and Serial Correlation
Another important factor that provokes positive serial correlation in return time series is
non-synchronous trading. If two stocks are trading in different frequencies, one could react
to news more quickly than the second. The lagged response of the second stock could
manifest as a positive serial correlation between the two returns. However, the effects of
non-trading may not be detectable in the returns of individual securities, but could be more
pronounced in portfolio returns (Lo & MacKinlay, 1990). Perry (1985) documents that non-
synchronous trading is not the only cause of correlation in daily market indices, but it
needs to be taken into account for the sake of the analysis.
The non-trading is typically associated with periods of low volatility but still present during
high fluctuation periods. In order to account for positive feedback trading alone, returns are
filtered using a linear autoregressive model LAR (p). The fitted returns have all the
autocorrelation induced by non-synchronous trading removed (Koutmos & Saidi, 2001).
4. Strategy Testing and Empirical Results 4.1. Trading Nature and Frequency
Since the Moroccan regulation does not allow for short selling in the Casablanca stock exchange, the analysis is conducted based on the actual trading mechanisms and for the hypothetical scenario where short selling is permitted to measure the practice’s impact on the different trading strategies. The strategies are also developed based on distinct trading frequencies. Each group of investors could trade on a daily basis or on a weekly basis. The daily trading allows the investor to decide on an action founded on the displayed closing price. The investor would act at the end of each trading day since the orders are assumed to be exercised instantaneous. In this method of trading, the investor obviously uses daily data (closing prices) as a basis for his decision making. For the weekly trades, the investors enters the markets on a given day of the week and keeps buying or selling (short selling) on that same day of every week. The weekly investor does not discard daily closing prices and bases his trades and forecasting tools on daily data. This investor also acts at the end of the trading exercise given the previous day’s price or the next day’s forecasted price (same day’s Sharpe ratio for smart money investors). Weekly trading based only on weekly data is also considered as one of the scenarios in this analysis. Traders would again trade on a given weekday, but would only look at
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
7
previous week’s closing price or next week’s forecasted price. This trading frequency is only considered for the case of positive feedback trading
4.2. Strategy Description
Passive Strategy This is simple buy and hold strategy that consists of keeping the index portfolio during the whole investment period. It involves entering a long position by buying the index on the first day (n) and selling it on the last day (t) of the study period. The transaction costs are accounted for in this paper and are considered to be equal to 0.22% of the transaction amount. Smart Money Investors’ Strategy This is an active trading strategy followed by rational investors that rely on stock fundamentals to make their investment decisions. The investors decide on the basis of the Sharpe ratio; they buy (sell) when the Sharpe ratio is higher (lower) than 1. For the case where short selling is allowed, investors buy when the Sharpe ratio is higher than 1 and short sell the index when the Sharpe ratio is lower than 1.
Sharperatio = E�R���� − ασ���
Where α is the risk free rate and is considered to be equal to 3.27% for the seven years period, and σ_(t-1) is the annualized standard deviation for the previous year’s return excluding transaction costs. Positive Feedback Trading Strategy This strategy is based on the trend chasing conduct. The investors base their decision on the price movement. So if the prices increase, investors buy the stocks and when the prices decrease they sell the stock. For the case of short selling, investors keep the same behavior in market increases and short sell the stocks in market declines. Simple Hybrid Strategy This strategy is adopted by positive feedback traders who integrate forecasting techniques in their trading conduct. Investors use GARCH in Mean (1,1) to model the volatility and a Linear Autoregressive Model to forecast the returns and the prices. Every year’s volatility and return forecast is based on GARCH-M estimates derived from the preceding three years’ data. Return equation: R4� = C� +C� ∗ R��� + C7 ∗ σ�� + ε Variance equation: σ�� = C8+C9 ∗ Residual���� + C= ∗ σ���� The residuals are assumed to follow a normal distribution and are computed based on actual returns and the forecasted values obtained from the return equation,
Residual� = R� −R4� The forecasted returns R4� and actual prices P� are used to predict the prices P�4 as the following: P�4 = R4� ∗ P��� +P���
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
8
If P���? ≥ P�, the feedback traders would buy the stock, and if P���? ≤ P� the feedback traders would sell the stock or short sell the stock in the relevant scenario. Complex Hybrid Strategy This strategy is also adopted by positive feedback traders who integrate GARCH modeling and forecasting in their trading. Investors estimate a GARCH model based on the Log Likelihood tests. The likelihood functions for the various sets of data are maximized using an Exponential GARCH (3,3,2) model with the Log of the variance in the mean equation. The auto regressive order for the linear mean equation is determined using the tests conducted in the data analysis section and is set to be p=1. The leptokurtic property of the time series reveals that a generalized error distribution (GED) is more suitable for the residuals’ estimation. Each year’s volatility and return forecast is based on E-GARCH-M estimates derived from the preceding three years’ data. Return equation: R4� = C� +C� ∗R��� + C7 ∗ Logσ��� + ε Variance equation:
Logσ��� = C8+C9 ∗ abs DResidual���Eσ���� F +C= ∗ abs DResidual���
Eσ���� F +CG ∗ abs DResidual��7Eσ��7� F
+ CH ∗ Residual���Eσ���� +CI ∗ Residual���
Eσ���� +C�' ∗ Logσ���� � + C�� ∗ Logσ���� �+ C�� ∗ Logσ���� �
The feedback trader strategy is as in the previous simple hybrid strategy.
5. Numerical results 5.1. Results based on trading frequency
Daily Trading The daily trading for all the strategies appears to be outperformed by the buy and hold strategy that leads to higher returns for the 2004-2010 period. This is mainly due to the high transaction costs incurred on a daily basis. The active traders however are better off during market downs since they benefit from smaller drawdown. Indeed, the buy and hold strategy suffered from a severe drawdown that other strategies did not go through.
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
9
Figure 1: Trading Strategies (daily) with No Short Selling The hypothetical short selling results in a quite different scenario; it allows traders to take advantage from the drawdown and succeed in beating the passive strategy. The herding traders profit from higher annualized returns and outpace other investors. The short selling does not boost smart money investors’ returns since the drawdown caused the Sharpe ratio to fall below 1. The SMI were inactive during and after the down period because their signaling incorporates yearly data and hence ended up with similar annualized return.
Figure 2: Trading Strategies (daily) with Short Selling The impact of short selling is more visible on herding strategies in the way that the annualized returns increase significantly accompanied with an increase in volatility. The passive strategy that appears to be the most attractive one at first sight reveals to procure the lowest Sharpe ratio since its returns are subject to severe fluctuations and a high level of volatility. The positive feedback trading on the other hand seems to be the most stable
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
250.00%
300.00%
350.00%
Passive Strategy
Positive Feedback
Trading
Hybrid Strategy
Complex Strategy
Smart Money
Investors strategy
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
250.00%
300.00%
350.00%
Passive Strategy
Positive Feedback
Trading SS
Hybrid Strategy SS
Complex Strategy SS
Smart Money
Investors strategy SS
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
10
strategy since it has the highest Sharpe ratio since it does not undergo from the sharp drawdown.
Passive
Strategy
Smart Money
Investors
Strategy
Positive
Feedback
Trading
Hybrid
Strategy
Complex
Hybrid
Strategy
Annualized
Return
No Short
Selling
32.28% 29.2% 26.69% 25.93% 25.18%
Short
Selling
32.28% 28.74% 45.36% 42.9% 41.35%
Volatility
No Short
Selling
39.28% 22.01% 11% 11.97% 11.55%
Short
Selling
39.28% 23.61% 15.43% 15.56% 15.53%
Sharpe Ratio
No Short
Selling
0.82 1.33 2.43 2.17 2.18
Short
Selling
0.82 1.22 2.94 2.76 2.66
Table 1: Annualized Return, Volatility, and Sharpe Ratio for Daily Trading
Weekly Trading (using daily data) The feedback strategies seem to be dominated by the passive and the smart money investors’ strategies on a weekly basis. The gap between the feedback strategies is intensified and the forecasting techniques appear to be more lucrative for weekly trading. The reduction of transaction costs allows for higher annualized returns. The short selling boosts those returns to increase even further and leads to the strategies having comparable payoffs in terms of the annualized returns.
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
11
Figure 3: Trading Strategies (weekly) with No Short Selling
Figure 4: Trading Strategies (weekly) with Short Selling The impact of the drawdown period is also visible on a weekly basis since it causes the passive strategy to be dominated. It leads to a stagnation of the smart money investors’ trading activity who still endures a high volatility level. The feedback trading outperforms the other strategies for both short selling and no short selling scenarios.
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
250.00%
300.00%
350.00%
Passive Strategy
Positive Feedback
Trading (W)
Hybrid Strategy (W)
Complex Hybrid
Strategy (W)
Smart Money
Investors Strategy (W)
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
250.00%
300.00%
350.00%
Passive Strategy
Positive Feedback
Trading SS (W)
Hybrid Strategy SS
(W)
Complex Hybrid
Strategy SS (W)
Smart Money
Investors Strategy
SS(W)
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
12
Smart Money
Investors
Strategy
Positive
Feedback
Trading
Hybrid
Strategy
Complex
Hybrid
Strategy
Annualized
Return
No Short
Selling
29.71% 23.28% 25.46% 22.11%
Short
Selling
29.71% 29.97% 33.80% 27.14%
Volatility
No Short
Selling
21.93% 11.35% 12.35% 11.78%
Short
Selling
23.54% 15.51% 15.55% 15.56%
Sharpe Ratio
No Short
Selling
1.35 2.05 2.06 1.88
Short
Selling
1.26 1.93 2.17 1.74
Table 2: Annualized Return, Volatility, and Sharpe Ratio for Weekly Trading
5.2. Results Based on Trading Nature
Positive Feedback Trading The positive feedback traders are better off trading on a daily basis during both market ups and downs. The short shelling enhances their annualized returns without having much of an effect on the volatility measures. It also widens the gap between the daily trading and weekly trading, showing that the drawdown is less for daily trading. The positive feedback trading is a stable and profitable strategy despite the high transaction costs the investors may incur.
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
13
Figure 5: Positive Feedback Trading with No Short Selling
Figure 6: Positive Feedback Trading with Short Selling
Daily Weekly
Annualized
Return
No Short
Selling
26.69% 23.28%
Short
Selling
45.36% 29.97%
No Short
Selling
11% 11.35%
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
1
11
8
23
5
35
2
46
9
58
6
70
3
82
0
93
7
10
54
11
71
12
88
14
05
15
22
16
39
Positive Feedback
Trading
Positive Feedback
Trading (W)
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
250.00%
300.00%
350.00%
1
11
8
23
5
35
2
46
9
58
6
70
3
82
0
93
7
10
54
11
71
12
88
14
05
15
22
16
39
Positive Feedback
Trading SS
Positive Feedback
Trading SS (W)
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
14
Volatility Short
Selling
15.43% 15.51%
Sharpe Ratio
No Short
Selling
2.43 2.05
Short
Selling
2.94 1.93
Table 3: Annualized Return, Volatility, and Sharpe Ratio for Positive Feedback Trading
Strategy Hybrid Strategy The hybrid strategy based on GARCH-M (1,1) forecasts is close to the positive feedback trading strategy in terms of outcome and profitability. They are also similar for daily and weekly trading and concerning the short selling effect. It is also a profitable strategy especially for the daily trades despite the fact that traders act based on forecasted prices instead of actual prices.
Figure 7: Hybrid Strategy with No Short Selling
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
1
12
7
25
3
37
9
50
5
63
1
75
7
88
3
10
09
11
35
12
61
13
87
15
13
16
39
Hybrid Strategy
Hybrid Strategy (W)
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
15
Figure 8: Hybrid Strategy with Short Selling
Daily Weekly
Annualized
Return
No Short
Selling
25.93% 25.46%
Short
Selling
42.9% 33.80%
Volatility
No Short
Selling
11.97% 12.35%
Short
Selling
15.56% 15.55%
Sharpe Ratio
No Short
Selling
2.17 2.06
Short
Selling
2.76 2.17
Table 4: Annualized Return, Volatility, and Sharpe Ratio for the Hybrid Strategy
Complex Hybrid Strategy The complex hybrid strategy using E-GARCH-M (3,3,2) is very similar to the hybrid strategy in terms of the gains’ characteristics, but the final outcome is lower compared to the other feedback strategies. This may be due to the fact that the forecasts incorporate with a higher precision the historical market dynamics, and result in more conservative prediction due to the sharp drawdown.
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
250.00%
300.00%
350.00%
1
13
6
27
1
40
6
54
1
67
6
81
1
94
6
10
81
12
16
13
51
14
86
16
21
Hybrid Strategy SS
Hybrid Strategy SS (W)
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
16
Figure 9: Complex Hybrid Strategy with No Short Selling
Figure 10: Complex Hybrid Strategy with Short Selling
Daily Weekly
Annualized
Return
No Short
Selling
25.18% 22.11%
Short
Selling
41.35% 27.14%
Volatility
No Short
Selling
11.55% 11.78%
Short
Selling
15.53% 15.56%
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
1
12
7
25
3
37
9
50
5
63
1
75
7
88
3
10
09
11
35
12
61
13
87
15
13
16
39
Complex Strategy
Complex Strategy (W)
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
250.00%
300.00%
350.00%
1
12
7
25
3
37
9
50
5
63
1
75
7
88
3
10
09
11
35
12
61
13
87
15
13
16
39
Complex Strategy SS
Complex Strategy SS
(W)
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
17
Sharpe Ratio
No Short
Selling
2.18 1.88
Short
Selling
2.66 1.74
Table 5: Annualized Return, Volatility, and Sharpe Ratio for the Complex Hybrid Strategy
5.3. Results Based on Day of Trade
Smart Money Investors Strategy When trading on a weekly basis, the smart money investors do not seem to be affected by the day they actually enter the market. The steadiness of their payoffs suggests that their trading behavior is less spontaneous and impulsive than other types of trading. The same remarks apply to the weekly trading with short selling for this category of investors.
Monday Tuesday Wednesday Thursday Friday
Annualized Return 29.74% 29.38% 29.71% 29.47% 29.31%
Volatility 22.02% 21.91% 21.93% 21.92% 21.91%
Sharpe Ratio 1.35 1.34 1.35 1.34 1.34
Table 6: Annualized Return, Volatility, and Sharpe Ratio Based on the Day of Trade for the
Smart Money Investors Strategy Positive Feedback Trading The positive feedback traders’ returns are highly influenced by the day they exercise their trades. The weekly outcomes of this strategy vary depending on the weekdays, with Wednesday being the best day to trade and Monday being the worst. The standard deviation of the Sharpe ratios is quite high (47.91%) which accounts for the importance of the entry day effect. The influence of timing on annualized returns remains identical when short selling is introduced and leads to higher variations in the Sharpe ratio translated by 67.11% standard deviation.
Monday Tuesday Wednesday Thursday Friday
Annualized Return 8.68% 13.93% 23.28% 16.70% 16.13%
Volatility 11.91% 11.88% 11.35% 11.63% 11.41%
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
18
Sharpe Ratio 0.73 1.17 2.05 1.44 1.41
Table 7: Annualized Return, Volatility, and Sharpe Ratio Based on the Day of Trade for the
Positive Feedback Trading Strategy
Figure 11: Weekly Positive Feedback Trading Using Daily Data Based on Day of Trade Hybrid Strategy The weekly outcomes of the hybrid strategy vary depending on the weekdays and the outcomes are different depending on the day of market entry, Wednesday is the best day to act and Monday is the worst with a Sharpe ratio lower than 1. The standard deviation of the Sharpe ratios is equal to 48.95% for the case of no short selling and 73.61% when short selling is simulated. The influence of timing on annualized returns remains identical when short selling is introduced; and Wednesday’s Sharpe ratio undergoes some improvement and Monday’s worsens.
Monday Tuesday Wednesday Thursday Friday
Annualized Return 10.50% 13.51% 25.46% 15.62% 17.38%
Volatility 14.27% 12.66% 12.35% 13.01% 12.70%
Sharpe Ratio 0.74 1.07 2.06 1.20 1.37
Table 8: Annualized Return, Volatility, and Sharpe Ratio Based on the Day of Trade for the Hybrid Strategy Using Daily Data
-40.00%
-20.00%
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
140.00%
160.00%
180.00%
Monday
Tuesday
Wednesday
Thursday
Friday
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
19
Figure 12: Weekly Hybrid Strategy Using Daily Data Based on Day of Trade Complex Hybrid Strategy The complex hybrid strategy also reveals Wednesdays as the best trading days and Mondays as the worst. The high importance of the entry and exit day is accounted for by the high standard deviation that is equal to 44.14% and 62.12% for the cases of no short selling and short selling respectively.
Monday Tuesday Wednesday Thursday Friday
Annualized Return 9.72% 11.07% 22.11% 14.06% 17.14%
Volatility 12.58% 12.44% 11.78% 12.11% 12.33%
Sharpe Ratio 0.77 0.89 1.88 1.16 1.39
Table 9: Annualized Return, Volatility, and Sharpe Ratio Based on the Day of Trade for the
Complex Hybrid Strategy Using Daily Data
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
Monday
Tuesday
Wednesday
Thursday
Friday
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
20
Figure 13: Weekly Complex Hybrid Strategy Using Daily Data Based on Day of Trade
5.4. Results Based on Data Frequency (for Positive Feedback Trading)
Effect of Data Frequency on the Return, Volatility, and Sharpe Ratio Weekly Positive feedback traders can either rely on daily data or on weekly data to exercise their trades. Using daily data earns higher returns than using weekly prices and is subject to a lower volatility of returns. It is also more profitable for positive feedback traders to act based on daily data since it yields higher Sharpe ratio.
Daily Data Weekly Data
Annualized Return
No Short Selling 23.28% 17.32%
Short Selling 29.97% 17.80%
Volatility
No Short Selling 11.35% 12.03%
Short Selling 15.51% 15.67%
Sharpe Ratio
No Short Selling 2.05 1.44
Short Selling 1.93 1.14
Table 10: Annualized Return, Volatility, and Sharpe Ratio for the Daily and Weekly Data (PFTS)
-40.00%
-20.00%
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
140.00%
160.00%
180.00%
Monday
Tuesday
Wednesday
Thursday
Friday
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
21
Effect of Data Frequency on the Day of Trade Using weekly data, the outcome of the positive feedback trading strategy still varies depending of the day of the trade. This change however, is smaller in magnitude than the variation induced when using daily prices. The day of the trade does not matter as much for weekly trading with weekly data, and results in a standard deviation of the Sharpe ratio of 36.94% in the absence of short selling. It is rather high for the short selling scenario case (56.79%), which is foreseeable since the short selling intensifies the market dynamics. When trading with weeks data, the volatility increases on average and the trading day with the highest outcome differs. The Sharpe ratio keeps increasing throughout the week and reaches its peak on Friday for both short selling and no short selling scenarios.
Monday Tuesday Wednesday Thursday Friday
Annualized Return
Daily Data 8.68% 13.91% 23.28% 16.70% 16.13%
Weekly Data 13.96% 7.11% 8.74% 16.65% 17.32%
Volatility
Daily Data 11.91% 11.88% 11.35% 11.63% 11.41%
Weekly Data 12.27% 11.60% 11.40% 12.00% 12.03%
Sharpe Ratio
Daily Data 0.73 1.17 2.05 1.44 1.41
Weekly Data 1.14 0.61 0.77 1.39 1.44
Table 18: Annualized Return, Volatility, and Sharpe Ratio Based on the Day of Trade for
the Daily and Weekly Data (PFTS)
6. Findings and Conclusion The study of the 2004 -2010 Casablanca Stock exchange main MADEX reveals crucial facts about the characteristics of the returns and the nature of trading conducts in the market place. The significance of the serial autocorrelation in the daily data indicates the plausible presence of positive feedback traders and trends chasers in the market. It also implies that this category of traders could influence market movements and induce negative autocorrelation in the stock returns, trigger mean reversion phenomena, and allow trend predictability. The analysis of the feedback traders or trend chasers’ strategies profitability leads to four main results. First, positive feedback trading on the daily basis beats other herding types of trading and reveals to be steadier than the simple buy and hold strategy whencomparing Sharpe ratios. It allows the traders to go through the intense 2008-2010
A. Charafi and M. Achkir Positive Feedback Trading and Stock Return Autocorrelation
Frontiers in Science and Engineering An International Journal Edited by Hassan II Academy of Science and Technology
22
drawdown and smooths the severity of the market fall. Positive feedback traders are subject to lower level of volatility even though they are pretty active in the market. The forecasting techniques however lead to inferior outcomes suggesting that there is a price to pay for information losses caused by forecasting. Secondly, the GARCH forecasting techniques provide superior outcomes for weekly trading using daily data. It allows the trend chasers to incorporate three years of data in their trading conduct. The forecasts seem to be more suitable for less frequent trades since the signaling is unchanged for the hybrid strategy and the complex hybrid strategy, and daily trades only increase transaction costs incurred by the investors. The daily herd trading however results in higher annualized returns when compared to weekly trading, while the volatility does not seem to be affected by the frequency of trades. Thirdly, short selling generally boosts the annualized returns and intensifies the volatility. The smart money investors are the only group that appears to be immunized from the influence of trading nature as well as frequency. Finally, the comparison of annualized returns and Sharpe ratio based on the day of trading for the weekly activity reveals that smart money investors are not affected by the day they enter or exit the market, while the herd traders are. In fact, Wednesdays are characterized by lower volatility levels when compared to other week days and yield higher returns. Mondays however are the worst days for feedback traders to act on the stock exchange as they are earning much lower returns and endure higher volatility levels. This phenomenon could be explained by the fact that the market goes through an adjustment stage in the beginning of the week and reaches the “steady state” by the mid-week.
References [1] Cutler, D.M., Poterba, J.M. & Summers, L.H. (1990). Speculative Dynamics and the Role of Feedback Traders. The American Economic Review, Vol.80 (2), pp. 63-68. [2] Shiller, R. J. (1984). Stock Price and Social Dynamics. Brookings Papers on Economic Activity, Vol. 2, pp. 457-498. [3] Sentana, E. and Wadhwani, S. (1992). Feedback traders and stock return autocorrelations: evidence from a century of daily data, The Economic Journal, Vol. 102, pp. 415-25. [4] Koutmos, G. (1997). Feedback Trading and the Autocorrelation Pattern of Stock Returns: Further Empirical Evidence. Journal of International Money and Finance, Vol 16, pp.625 – 636. [5] Koutmos, G., & Saidi, R. (2001). Positive Feedback Trading in Emerging Capital Markets. Journal of Applied Financial Economics, Vol. 19 (24), pp. 291-297. [6] Squalli, N. (2006, November 22nd). Colorado: L’Effet Moutonnier. L’Economiste. [7] Drissi El Bouzaidi, O. (2006, February, 10th). Le Marché Boursier Prend Près de 30 % : Bulle Spéculative ou Pas ? LavieEco. [8] Nhaili, S. (2009, January, 16th). Perte de Confiance et Manque de Visibilité à la Bourse de Casa. LavieEco. [9] Lo, A. W., & MacKinlay A. C. (1991). An Econometric Analysis of Non-Synchronous Trading. Journal of Econometrics, Vol.45 (1-2), pp. 181-211. [10] Perry, P. R. (1985). Portfolio Serial Correlation and Non-Synchronous Trading. The Journal of Financial and Quantitative Analysis, Vol. 20 (5), pp. 517-523
top related